Partial Discharge Testing of Random Wound Stators ... - IEEE Xplore

2009 IEEE Electrical Insulation Conference, Montreal, QC, Canada, 31 May - 3 June 2009
Partial Discharge Testing of Random Wound Stators
During Short Risetime Voltage Surges
G.C. Stone (FIEEE) and I. Culbert (SMIEEE)
Iris Power LP
3110 American Drive
Mississauga, ON, Canada L4V 1T2
Abstract: Modern inverter fed motors often produce short
risetime, high magnitude voltage surges that may lead to partial
discharge in the random wound stator windings. Detection of PD
during such voltage surges can not be done with the usual PD
detectors described by ASTM D1868. Instead a detector that can
separate PD pulse currents from a short risetime voltage surge is
required. This paper describes a PD detector that can measure
PD pulses when the applied voltage is a voltage surge with a
risetime as short as 100 ns. PD data has been collected from
dozens of stators, many of which had known insulation problems.
In most cases the PDIV easily met the requirements of a new IEC
technical specification that sets the partial discharge inception
voltages for random wound stators. Since most of these stators
had obvious insulation system deficiencies, it seems that the
PDIV levels set by IEC TS 60034-18-41 may be too low.
(PD) [1-3]. Since the magnet wire insulation is primarily
made from organic materials, if sufficient PD activity occurs,
the insulation deteriorates and eventually fails. Thus it is
usually desired that no PD occurs during operation of the
motors. IEC TS 60034-18-41 sets the minimum voltages that
PD may occur (Table 1).
This paper discusses how the PD can be measured during
short risetime voltage surges, and presents some PD data
measured on many stators during short risetime surges.
I. INTRODUCTION
Variable speed drives of the voltage source, pulse width
modulation type have become one of the most popular types
of motor drives. Such drives can produce switching voltage
surges with a risetime as short as 100 ns, and with thousands
of surges per second [1]. There is anecdotal evidence that the
large number of voltage surges from such inverter fed drives
(IFDs) may lead to the deterioration and eventual failure of the
insulation in low voltage (less than 1000 V) motor stator
windings [1-4]. Surges over 3000 V and risetimes as short as
100 ns have been measured on motors that are nominally rated
up to 690 Vrms (phase to phase) [3].
The stator windings are “random wound” in most motors rated
less than 1000 V. This implies that the magnet wire connected
to the phase terminals in one phase, may be adjacent to
magnet wire in another phase or a magnet wire near the
neutral. In addition, for short risetime pulses, the voltage is
not uniformly distributed from the phase terminal down to
neutral. Instead, most of the voltage is dropped across the
turns at the line –end of the winding [2]. As a result,
unusually high voltages can occur across adjacent turns of
magnet wire; between magnet wires in different phases; and
between the magnet wire and the grounded stator slot.
The electrical stresses may be high due to these voltages, the
relatively small diameter of the magnet wire and the thinness
of the magnet wire insulation. As shown in Figure 1, if there
are any air spaces between two adjacent wires and/or ground,
the stress may be high enough to initiate partial discharge
Fig. 1: PD occurring in air space between magnet wires at high stress.
II. MEASURING PD DURING VOLTAGES SURGES
ASTM D1868 and IEC 60270 describe the measurement of
PD in an insulation system exposed to power frequency
voltage. Such conventional PD detectors usually have a high
voltage capacitor to reduce the high AC voltage to the
millivolt range, yet pass the PD signals to an oscilloscope or
other recording apparatus virtually unattenuated. However,
these conventional methods to measure the pulse current
associated with each partial discharge cannot be used with fast
risetime voltage surges because the PD pulse has much the
same frequency content as the surge. Thus the standard high
pass filter characteristics of a PD detection capacitor will
apply several hundreds of volts to the PD measurement
electronics during each surge, destroying the electronics.
A specialized PD measuring system is needed which can
record the PD pulses during a surge. This involves
suppressing the surge voltage (up to 3 or 4 kV) to less than a
volt, while passing the PD pulse with little or no attenuation.
IEC TS61934 was recently issued that addresses this
measurement problem. The TS describes a number of
measurement methods. All but one are classed as working in
the VHF or UHF frequency range, since the PD pulses are
978-1-4244-3917-1/09/$25.00 ©2009 IEEE
188

detected in the hundreds of megahertz range due to the PD
pulse risetime of only about 1 ns [5]. Although high order
high pass filters have been used to suppress the voltage surge,
these do not seem to be effective in suppressing the residual of
the voltage surge to about the same magnitude as the PD,
when the surge risetime is less than 200 ns or so.
TABLE I
RPDIV MEASURED WITH A SURGE TESTER FOR A 480 V RATED MOTOR STATOR
FED FROM A 2 LEVEL INVERTER
Stress
Category
Peak
V/DC
Bus
RPDIV (V p-p )
Phase to Phase to
Phase Ground
Turn
to
Turn
Benign 1.1 1853 1297 1297
Moderate 1.5 2527 1769 1769
Severe 2.0 3370 2359 2359
Extreme 2.5 4212 2948 2948
The only devices that seem to have sufficient suppression
capability for the risetimes typical of voltage source invertors
use either a very small UHF antenna near the winding or a
directional electromagnetic coupler. Small UHF antennae
seem to be relatively insensitive to PD. However, 10 years of
experience has shown that a directional electromagnetic
coupler technique has good sensitivity to PD, while adequately
suppressing high voltage surges with risetimes as short as 100
ns [6]. An instrument called PDAlert uses the directional
electromagnetic coupler approach (Figure 2). The PD output
of the instrument is displayed on a digital oscilloscope, as is a
divided version of the voltage surge. As with all VHF and
UHF PD detectors, the PD magnitudes should not be
expressed in pC [5], thus the PD magnitude is measured in
terms of mV. A sensitivity check is made, using the method
described in IEC TS 61934, or by using a twisted pair of
magnet wire with a low PDIV.
Figure 3 shows a single PD pulse recorded in an off-line test
of a 5 HP motor using a Baker Model D12000 surge tester.
The surge tester has a surge risetime of about 100 ns, at the
motor terminals. Depending on the load, the risetime with the
D12000 is between 100 ns and 400 ns. The Baker D12R and a
modern PJ tester have also been shown to produce short
risetime outputs into stators. The high magnitude, high
frequency output on the lower trace is due to PD. It is
oscillatory (an artifact of the measurement system), and has a
risetime of about 5 ns. It is evident on the lower trace that
there is still some of the residual surge that comes through the
instrument. However it has a lower frequency content and
linearly increases in magnitude as the surge voltage increases.
These characteristics make it easy to distinguish the surge
residual from the PD pulse. The surge voltage has been
reduced by over 66 db (2000 times).
PD-free is defined in IEC TS 61934 as having a repetitive
partial discharge inception voltage (RPDIV) higher than a
specified voltage, with a surge risetime similar to that which is
expected in service. The RPDIV is measured by slowly
increasing the voltage from the surge tester, which outputs one
surge per second. The surge residual voltage gradually
increases from 0. The PD inception voltage (PDIV) is
determined when the high frequency PD pulse burst is first
noted on the detector output trace on the oscilloscope.
However, at this voltage, PD pulses will often not occur on
repeat surges. By increasing the voltage (usually by 50 to 300
V), the PD occurs on most of the applied surges, and this is
termed the RPDIV (according to IEC TS 61934 the lowest
voltage at which a PD occurs during 50% of the surges). After
one hour or so of training, factory floor technicians have been
able to reliably measure the RPDIV.
The magnitude of the surge residual in Figure 3 depends on
the surge risetime that actually is impressed on the stator
winding (the shorter the risetime, the higher is the residual).
For most of the motors tested to date, the residual had a
magnitude between 500 and 1500 mV at the RPDIV. In
comparison, the PD magnitude at the RPDIV depends on the
effective load capacitance of the winding and the distance the
PD pulse has to travel to the PD instrument. For most of the
motors tested to date, the PD magnitude was between 300 mV
and 1000 mV. The PD magnitude increases dramatically as
the surge voltage increases above the RPDIV [7].
Scope
Surge generator
PD detector
Stator
Fig. 2. RPDIV measurement of a small motor stator (right side). The voltage
surges come from a Baker D12000.
Fig. 3. PD (lower trace) recorded during a single surge (upper trace) from a
surge tester applied to a small stator. The time base is 50ns.div. The surge
voltage is 1000 V/div while the PD is 1 V/div. The low frequency transient at
the beginning of the PD waveform is the residual from the surge. The high
frequency oscillation in the middle of the lower trace is the PD.
189

III. EXPERIMENTAL PD DATA
The RPDIV has now been measured on over 100 stators, and
the data formally recorded from 39 machines. Table 2 shows a
summary of the RPDIV collected by ourselves using either a
D12R or a D12000 surge tester. The test objects were all
stator windings, rated from 1 HP to 6400 HP and from 380 V
to 690 V. Almost all were random wound stators. Most of the
measurements in Table 2 occurred because the motor OEM or
the enduser had experienced premature stator winding
insulation failures, and PD due to IFD voltage surges was one
of the possible failure mechanisms.
The measured surge risetime was from 150 to 400 ns, and the
surge pulse width for most stators ranged from 10 to 40 μs,
with a slow fall-time. Due to the long width and slow fall, and
since the opposite polarity undershoot was always less than
25% of the peak surge magnitude, the RPDIV is measured
from 0V to the peak, as suggested in IEC TS 60034-18-41
(Table B.2, Column 6, Note b). This is valid for the turn to
turn RPDIV. Example minimum RPDIVs are shown in Table
1. The phase to ground RPDIV may be as much as 25%
higher than shown in Table 2.
The second last column shown in Table 2 is the lowest RPDIV
measured on all phases and with all possible test connections
in 39 stators. The PD may be occurring between turns,
between a phase and ground, or in the endwinding between
phases. If the RPDIV is indicated with “>”, the RPDIV was
not measured since the motor owner did not want to exceed
the indicated voltage. The measured RPDIVs can be
compared the minimum required RPDIV for a “moderate” and
“severe” surge environment in columns 4 and 5, as defined in
the IEC TS.
TABLE II
MEASURED PHASE TO GROUND AND TURN TO TURN RPDIV COMPARED TO THE MINIMUM REQUIRED RPDIV FOR THE MODERATE
AND SEVERE STRESS CATEGORIES.
Rating DC Bus Min. Turn or Ground Measured
Condition
RPDIV (Volts) Min. PDIV
HP Volts Volts Moderate Severe Volts
1 440 590 1593 2124 1900
10 440 590 1593 2124 2200
20 440 590 1593 2124 1800
650 440 590 1593 2124 1250
650 440 590 1593 2124 2080
1100 690 975 2632 3510 1352 Failed
170 690 975 2632 3510 983 Failed
6400 1287 1600 Failed (tr 500 ns)
880 600 850 2295 3060 2200
300 600 850 2295 3060 2200 Failed
300 600 850 2295 3060 >2500
300 600 850 2295 3060 >2500
300 600 850 2295 3060 >2500
300 600 850 2295 3060 >2500
880 600 850 2295 3060 1890 Cast, cracks
950 600 850 2295 3060 1240 No phase insulation
880 600 850 2295 3060 >2600
950 600 850 2295 3060 >2400
950 600 850 2295 3060 >2600
880 600 850 2295 3060 2320 Contaminated
1000 690 975 2632 3510 >2000
1000 690 975 2632 3510 2600
1000 690 975 2632 3510 >2000
17 415 560 1512 2016 2960
17 415 560 1512 2016 2160 No phase insulation
35 415 560 1512 2016 2640
35 380 513 1385 1847 2480
35 380 513 1385 1847 2960
17 380 513 1385 1847 2240
17 380 513 1385 1847 2560
17 380 513 1385 1847 2320
36 380 513 1385 1847 2720
17 380 513 1385 1847 1400 Failed
17 380 513 1385 1847 1900 Poor impregnation
17 380 513 1385 1847 1400 Poor impregnation
350 440 590 1593 2124 >2000 Similar motors failed
350 440 590 1593 2124 >2000 Similar motors failed
350 440 590 1593 2124 2700 Similar motors failed
350 440 590 1593 2124 3100 Similar motors failed
190

Some observations on these results are:
• 4 stators with known problems exceeded the
“severe” level, and thus the level set for the severe
level seems to be too low for these stators since the
“bad” windings still would have passed.
• 9 of the 39 stators with known problems had
RPDIVs that exceeded the moderate level. Thus
these motors would have passed a moderate severity
RPDIV test, even with the known problems. Of
course none of the motors were originally qualified
to the new IEC TS, since most of the motors predate
the TS. Thus we do not know the stress category
the motors were designed for.
• The RPDIVs range from 1000 V to 3000 V (for all
voltage ratings). For motors of one voltage rating
(say 440 V), the RPDIV ranged from 1250 V to
3100 V. Clearly there is a great difference in
RPDIVs between stators, and presumably a great
difference in design, manufacturing and
impregnation.
Data published by GE on 18 motors of the same rating from
6 manufacturers also showed high variability in the RPDIV –
ranging from 1600 V to 3200 V [8]. GE also reported that in
factory tests on 2088 motors rated 440V, the PD ranged from
2950 V to 3690 V [8]. All of these motors easily passed the
RPDIV acceptance requirement for a “severe” insulation
environment.
IV. CONCLUSIONS
1. A new IEC technical specification has been issued which
sets the minimum PD inception voltages for inverter fed
motors, based on the expected severity of the voltage surges
at the motor terminals.
2. PD measurement equipment that is capable of separating
PD from the relatively large applied voltage surges has been
in commercial use for 10 years. This PD detector will
separate PD from the surges for surge risetimes as short as
100 ns – which is the risetime of surges that can occur from
IFDs. At least two OEMs use the detector routinely on the
factory floor to ensure production motors have a suitably
high RPDIV.
3. Separating the PD from the various sources (turn, ground
and interphasal insulation) is not trivial. In addition to
measuring the PD under surges, it is useful to also measure
the PD with 50/60 Hz AC voltage applied to the windings.
This will help to separate turn to turn PD (which only occurs
with surge voltages) from ground and interphasal PD (AC
will only excite ground and interphasal PD).
4. Production motor acceptance testing by others shows the
RPDIV easily exceeds the IEC TS 60034-18-41 requirements
for Type I insulation systems, thus indicating that perhaps the
minimum RPDIV levels that have been established are too
low. This is further supported by the 39 stators reported
here, since some of the motors which had known
deficiencies, also passed the current minimum levels.
REFERENCES
[1] Persson, E. ‘Transient Effects in Applications of PWM Inverters to
Induction Motors’, IEEE Trans IAS, Sept 1992, p1095.
[2] Stone, G.C., et al, “Electrical Insulation for Rotating Machines”, Wiley
IEEE Press, 2004
[3] Stone, G.C., Campbell, S., Tetreault, S. “Invertor Fed Drives: Which
Motor Stators are at Risk”, IEEE Industry Applications Magazine, Sept
2000, pp17.
[4] Wheeler, J., “Effects of Converter Pulses on Electrical Insulation – The
New IEC Technical Specification”, IEEE Electrical Insulation
Magazine, Mar 2005, p22.
[5] S. Coen et al, “Sensitivity to UHF PD in Power Transformers”, IEEE
Trans DEIS, Dec 2008, p1553.
[6] Stone, G.C., Campbell, S., Susnik, M. “New Tools to Determine the
Vulnerability of Stator Windings to Voltage Surges from IFDs”, Proc.
IEEE Electrical Insulation Conference, Cincinnati, October 1999,
p149.
[7] Fenger, M., Stone, G.C., “Progress in Understanding the Nature of PD
in Stators Created by Inverter Drives”, Proc. IEEE Electrical Insulation
Conference, Sept 2003, p363.
[8] Bogh, D, et al, “Partial Discharge Inception Testing on Low Voltage
Motors”, IEEE Trans IA, Jan 2006, p148.
191